Anti-crosstalk high-fidelity state discrimination for superconducting qubits

TL;DR

This paper introduces a shallow neural network-based digital signal processing system to improve qubit state discrimination in superconducting quantum systems, significantly reducing crosstalk-induced errors and enhancing measurement fidelity.

Contribution

It presents a novel neural network approach integrated into DSP for qubit measurement, achieving near-complete crosstalk error reduction in multi-qubit superconducting devices.

Findings

Crosstalk-induced readout error decreased by 100%

Neural network optimization improved measurement fidelity

Effective in a 6-qubit superconducting quantum chip

Abstract

Measurement for qubits plays a key role in quantum computation. Current methods for classifying states of single qubit in a superconducting multi-qubit system produce fidelities lower than expected due to the existence of crosstalk, especially in case of frequency crowding. Here, We make the digital signal processing (DSP) system used in measurement into a shallow neural network and train it to be an optimal classifier to reduce the impact of crosstalk. The experiment result shows the crosstalk-induced readout error deceased by 100% after a 3-second optimization applied on the 6-qubit superconducting quantum chip.